Subscribe to The Podcast by KevinMD. Watch on YouTube. Catch up on old episodes!
Dermatologist Michael Sherling discusses his KevinMD article, “Where physicians need to implement AI first.” In this episode, Michael examines the potential of artificial intelligence to alleviate administrative burdens in health care, such as documentation management and patient collaboration. He emphasizes the importance of prioritizing AI tools that deliver high value with minimal risk, starting with patient-facing applications that streamline appointment scheduling and routine inquiries. Michael also explores the challenges of integrating AI into clinical workflows, the risks of biased or poor-quality data, and the necessity of maintaining a personal touch in patient care. Additionally, he offers actionable strategies for physicians to effectively implement AI solutions that enhance practice efficiency, reduce staff burnout, and ultimately allow doctors to focus more on patient interactions.
Our presenting sponsor is DAX Copilot by Microsoft.
DAX Copilot, by Microsoft, is your AI assistant for automated clinical documentation and workflows. DAX Copilot allows physicians to do more with less and turn their words into a powerful productivity tool. DAX Copilot automates clinical documentation—making it available in the EHR within minutes—and clinical workflows, including referral letters, after-visit summaries, style and formatting customizations, and more.
70 percent of physicians who use DAX Copilot say it improves their work-life balance while reducing feelings of burnout and fatigue. Patients love it too! 93 percent of patients say their physician is more personable and conversational, and 75 percent of physicians say it improves patient experiences.
Discover AI-powered solutions for clinical documentation and workflows. Click here to see a 12-minute DAX Copilot demo.
VISIT SPONSOR → https://aka.ms/kevinmd
SUBSCRIBE TO THE PODCAST → https://www.kevinmd.com/podcast
RECOMMENDED BY KEVINMD → https://www.kevinmd.com/recommended
GET CME FOR THIS EPISODE → https://www.kevinmd.com/cme
I’m partnering with Learner+ to offer clinicians access to an AI-powered reflective portfolio that rewards CME/CE credits from meaningful reflections. Find out more: https://www.kevinmd.com/learnerplus
Transcript
Kevin Pho: Hi, and welcome to the show. Subscribe at KevinMD.com/podcast. Today, we welcome Michael Sherling. He’s a dermatologist, and today’s KevinMD article is “Where physicians need to implement AI first.” Michael, welcome to the show.
Michael Sherling: Thanks so much, Kevin. It’s great to be here. I’m really excited.
Kevin Pho: All right. So tell us a little bit about yourself briefly and how you got interested in talking about AI and writing this article for KevinMD.
Michael Sherling: Happy to. Like many of your listeners, my path to medicine started out in a traditional way, taking basic science classes in the first year of med school. But it turns out serendipity has a way of following me, and I ran into a classmate who was auditing a course, how to start a high-tech company at the business school. I kind of took that and audited it at night while I was a med student by day and really just fell in love with being an entrepreneur. I became hooked. I learned how to write a business plan, cultivate an idea, and how to pitch it. I took that creativity with me and did an MBA during med school. It was really one of the best decisions that I made.
At the time though, as someone coming from med school, you can’t really contribute in business school in a meaningful way. You don’t have a history of working in the business world to really contribute to conversations. But I banked that and waited for my opportunity. I graduated, went on to residency in dermatology, and started to see the pain points in health care and how health care tech is really designed for the consumers of information rather than the creators of information, which are the doctors and the patient. I moved to Florida. I married my med school sweetheart, who is from Florida, and went into private practice.
When serendipity hit me again, I ran into a patient who is now my co-founder, Dan King. He just wanted a skin check, but he had written on his intake sheet that he was a serial entrepreneur. He had founded a company called Blackboard, which is the e-learning company that he took public. He started that in college and then took that public. So he was in between companies. He was sitting in my exam room, naked, with a blue gown on. I was supposed to give him a skin check, but I started peppering him with questions about how we can make a health care technology work for physicians. And he was like, “I’m cold and I’m naked. Can you please finish the exam exam?” So we did. Then I took him out to lunch, and we literally just started the company over lunch. We started writing different designs on napkins and created a business plan. For 15 years, it’s been growing ever since. The company’s called ModMed, and it delivers health care technology solutions to physicians and their staff.
Kevin Pho: All right. So, this article specifically talks about AI, and of course, that’s such a huge intersection now. I’ve had several podcasts talking about the influence AI has on medicine, but with your background in dermatology, I’m interested in hearing your perspective, because AI has been used in dermatology even before this whole OpenAI ChatGPT craze. So, tell us about the article itself, and from your perspective as a dermatologist, how AI is being used in medicine today.
Michael Sherling: Yeah. I think AI is obviously getting a lot of buzz in health care, and it has so much promise for all of us. When you reference things around what AI can do, there are certain models that look at different moles and predict whether they should be biopsied or what the diagnosis could be based off of pictures. I’m really not talking about that AI application. I think doctors like to practice, and they feel good about their clinical skills. I think what they need help with, more than anything, is curing the drudgery of health care—the things that burn us out all the time, like answering nonclinical questions, filing electronic passes, and charting. These are the types of things that doctors really want AI for. So, in the KevinMD article, I wanted to share my thought process on how we can introduce AI capabilities to users so that we can get comfortable with it. I think, in general, health care as a whole industry is a little bit late to the game on adopting technology. I wanted to outline in the article what is hype and what is helpful, and how AI works so that doctors can understand how best to use it.
Kevin Pho: All right, so tell us about some of the tools that have, of course, the highest value and the least risk to physicians from your perspective.
Michael Sherling: Absolutely. I think it starts with a patient. Tools that can work on helping draft interclinical or nonclinical communication, like directions to an office or what time you start, or simple questions, are perfect for AI. So, in a patient engagement tool that answers questions or helps assist in scheduling patients for the right appointment, that’s something easy to do. Staff take a lot of time and resources to hire people at the front whose job is to greet the patient in front of them, but they’re tethered to a computer answering messages for all the other patients. So, I think that’s a win-win.
I think in the clinic, for the doctors listening in, nobody went to medical school to write a note in an EMR. We want to treat patients. Ambient listening technologies that are out there show great promise, but it isn’t just about writing the note, right? It’s about all the downstream actions that the clinician still needs to do, like prescribe medication, order labs, order the pathology requisition, deliver education notes to the patient, and fill out the encounter form with the ICD-10 code. I think it’s a start, but we have to really think about saving time, not just using shiny AI technology.
Kevin Pho: So let’s talk about those two individually. For the first one, in terms of answering general logistics questions—like do you have any appointments available or how to get to your clinic—are you talking about some type of AI chatbot that patients can use on the front end of a practice website, for instance?
Michael Sherling: Yeah. I think there are definitely many ways it could work. We have a patient engagement tool at ModMed, and we’re exploring how to train the model to answer the most common questions that are not clinical that patients may have. That would get served up to a front desk person, and then they could decide if they like what it generated. If they do, awesome; if they don’t, they can change it and tweak it. But why do we have to craft everything ourselves from scratch? I think that has tremendous value.
Kevin Pho: All right, and of course, ambient AI is all the rage right now. I talk about that all the time on my podcast. Full disclosure: this podcast’s presenting sponsor is an ambient AI scribe company. Now, going beyond just a note, you’re talking about having it order tasks, almost like AI agents, based on what’s said in the chart. Talk to me about that next evolution you see.
Michael Sherling: Yeah, absolutely. I think AI is just AI tools, right? There are large language models that can classify and summarize what a doctor and a patient are saying, and there are ways of predicting different elements. But what I’m talking about is predicting structured data. If you can train an AI toolset not just on the conversations between the doctor and the patient, but also on the different medications that could be prescribed for a given diagnosis, the exam findings that would be present, or the types of surgeries typically ordered, then we can automate downstream tasks. Now, the doctor is still the decision maker—the AI isn’t doing anything the doctor isn’t saying—but it will queue up everything the doctor said and present it as, “Hey doctor, you’re the human in the loop. You said you wanted to prescribe doxycycline for this acne. You said you wanted to order a CBC and diff. Here they are. Press this one button, and it all goes into your note. Do you want this?” Who wouldn’t want that? Why do we have to recreate everything sometimes?
Kevin Pho: So it’s almost like some type of predictive tool. Like whenever you type a Google search, Google will predict what you’re going to search for. It’s up to you to hit enter or not.
Michael Sherling: Exactly, and that prediction is based on data, right? We build specialty electronic health records in dermatology, ophthalmology, orthopedics, so we know from millions of notes what the patterns are in terms of diagnoses, prescriptions, and surgeries. We just want to serve that knowledge back to the practicing physician: “Hey, you said this. Do you want this?” And instead of just writing a note, it writes the note, the prescriptions, the encounter form, and all the downstream actions. I think that’s really the time and the savings that we’re looking for.
If you think about where we were before all of this technology craze with meaningful use, we had medical assistants and receptionists, but I feel like over the last 15 years, the job descriptions changed. They’re more like assistants to the EMR and practice management system. I think AI can recoup a lot of that back and let people treat patients, see patients, assist the doctor in setting up the surgical tray, and look a patient in the eye and greet them. I think that’s what we’re looking for.
Kevin Pho: And that predictive tool you’re mentioning—is that technology available yet?
Michael Sherling: Yeah. So it’s something we’re building. We’re in beta right now in dermatology. We’re building out the other specialties this year, including orthopedics, ENT, podiatry, ophthalmology, and the rest of our specialties by the end of the year.
Kevin Pho: Where else do you see AI helping with the drudgery of medical care? I always mention the inbox. The inbox is the bane of a lot of physicians, especially for me in primary care. Specifically with the inbox, how can AI help us with that?
Michael Sherling: Exactly. It’s all about data, right? If we’re just trying to respond to a message—someone sending me a message who just wants a response—then you’d train an AI model on the messages you have. But I think we want to be a bit careful, because clinical messages carry a higher risk than nonclinical messages. That might be an area we take on a little bit later. Certainly, we could summarize what a doctor wants to say and have the doctor approve it.
If there are different clinical areas that are the bane of our existence, it’s prior authorization. We all hate prior auth. I think prior auth is similar to ambient listening, where you’re talking about what the patient tried and failed, what the severity of the illness is, and so on. That could be captured, collected, and predicted. That would be easy, whether it’s a prior auth for medication or for a procedure. I think we all spend time and money on that, and nobody would be sad if we didn’t have to.
I think faxing—why medicine is still focused on faxing is crazy, but that’s how a lot of us communicate. Why do we have to pay somebody, or why does that person have to spend time reading the fax, mapping it to the patient’s name, and putting it into the category of prescription refill, outside medical records, or pathology results? That is something AI could do, and we’re actively working on it.
On the revenue cycle management side, there are plenty of opportunities for AI disruption, like following a claim all the way along. Why do so many people have to touch the claim? Why do we have to generate medical necessity ourselves? Why can’t we just read the note and generate the medical necessity or supporting documentation? I think there are lots of opportunities on the revenue cycle side.
You look at practices today: they’re burned out, their staff is burned out, there are medical shortages, and you can’t hire the right people even if you wanted to. A lot of scribes today want to go to medical school—that’s their path to medical school. Maybe it wasn’t when we were there, but these are the kinds of things that I think medicine can’t get enough of. I look forward to the time when technology can actually deliver on its promise, which is saving doctors time and letting them enjoy treating patients, which is what we want.
Kevin Pho: So, of course, you’re a physician and also in the health startup phase, and you’re in an AI-centric company as well. I think back when electronic medical records came out, there weren’t a lot of physicians guiding that innovation. Hopefully with AI, you see more and more physicians guiding the information, and like you said before, you want these tools from more of a creator standpoint. Are you seeing more physicians today involved in AI-related tech startups?
Michael Sherling: Absolutely. I went to one of my reunions—I did an internship at the Brigham and Women’s Hospital—and I was pleasantly surprised to see so many physicians taking leadership roles in AI and different areas of medicine. I think we’re the right stewards of this, so I’m really excited about that. Our model is to hire physicians too. I have 21 physicians under me in different specialties. We teach them how to code in JavaScript and XML, and they build medical knowledge bases. They’re the ones working with data scientists, training the AI models as well, and they use their practices as a lab. I think that’s the right way to do it. You want physicians in the center, deciding how to use the technology and what it’s good for. I think all of us are more focused on giving time back to doctors rather than figuring out how to make a diagnosis. We know how to make a diagnosis; that’s not our problem. Our problem is bandwidth and spending time with our patients, and also getting out of clinic at five and spending time with our families. That’s what we want.
Kevin Pho: So what do you see as the near-term trends when it comes to that intersection between AI and health care? What do we have to look forward to in the coming year?
Michael Sherling: I think it’s going to be an exciting year. There’s a lot of promise. With every new technology, there will be some learnings along the way. We have to figure out the balance of what we want AI to predict and what we want humans to do. We’ll figure out that balance, but I think the ambient listening companies are going to get it right. I think doctors will start to see the promise of technology. We’ll also start working on clinical and nonclinical workflows for notifying patients of results, prior authorization, faxing, and the inbox that you hate so much. Those are areas we’ll start to get help with, which is what we want.
For people disillusioned with technology, who have spent more time and money hiring more people to support implementations of health care technology, I think you’ll be pleasantly surprised when technology works for you instead of the other way around.
Kevin Pho: We’re talking to Michael Sherling. He’s a dermatologist, and today’s KevinMD article is “Where physicians need to implement AI first.” Michael, let’s end with some take-home messages that you want to leave with the KevinMD audience.
Michael Sherling: Absolutely. I’d say that as a designer of health care software, clicks matter. The technology has to reduce clicks to save time. Shiny technology without time savings won’t be used. That’s something I think other people and technologists should hear: it has to save the doctor time. You have to eat your own dog food. When we design software, we use the products we create. Not everything you bring to life will work exactly as you want it to, so you have to test it yourself to truly advocate for the medical community.
Finally, be persistent. Don’t accept “no” at face value. If there’s something you want to achieve out there, there are ways of doing it. Believe in yourself and your team, and achieve it.
Kevin Pho: Michael, thank you so much for sharing your perspective and insight, and thanks again for coming on the show.
Michael Sherling: Thank you.
